Durvalumab for treating limited-stage small-cell lung cancer after platinum-based chemoradiotherapy

Small-cell lung cancer (SCLC) is an aggressive type of lung cancer that grows rapidly and spreads quickly to other areas of the body. In most people diagnosed with SCLC, the cancer has already spread (metastasised). But in about 30% of people, the cancer is contained in a single area that can be treated with radiotherapy. This is known as limited-stage SCLC (LS‑SCLC). The patient organisation's submission explained that a diagnosis of SCLC is devastating. People are often symptomatic at presentation and symptoms such as breathlessness, coughing up blood, hoarseness, chest pain and weight loss are distressing for both people with SCLC and their families. The patient organisation explained that SCLC is a rapidly progressing disease with a poor prognosis. The professional organisations' submissions explained there have been limited advances in the treatment of LS‑SCLC for over 20 years, and there are no maintenance treatments available after chemotherapy and radiotherapy. People with the condition and healthcare professionals would welcome more options early in the treatment pathway because of the poor prognosis. The committee agreed there is an unmet need for new and effective treatments for LS‑SCLC, particularly earlier in the treatment pathway.

NICE's guideline on diagnosis and management of lung cancer provides recommendations for the first-line treatment of LS‑SCLC. These include 4 to 6 cycles of platinum-based combination chemotherapy alongside radiotherapy, known as concurrent chemoradiotherapy (cCRT). People who are too unwell to have cCRT may have radiotherapy after chemotherapy, known as sequential chemoradiotherapy (sCRT). Surgery may also be an option for some people with early (stage 1 to 2) SCLC, but surgery is not possible for most people with LS‑SCLC. Prophylactic cranial irradiation (radiotherapy to prevent or reduce the risk of cancer cells spreading to the brain) may also be used after chemotherapy and radiotherapy. After first-line treatments, people are actively monitored for cancer recurrence. This involves outpatient appointments and regular imaging (scans). Durvalumab is positioned as an addition to active monitoring when the cancer has not progressed after platinum-based chemoradiotherapy. So, the comparator is active monitoring alone. If the cancer recurs after first-line treatment, second-line treatment options include further treatment with a chemotherapy regimen, with or without an immunotherapy such as atezolizumab (see NICE's technology appraisal guidance on atezolizumab with carboplatin and etoposide for untreated extensive-stage small-cell lung cancer) or durvalumab (see NICE's technology appraisal guidance on durvalumab with etoposide and either carboplatin or cisplatin for untreated extensive-stage small-cell lung cancer). Palliative radiotherapy may also be offered.

The clinical evidence for durvalumab in LS‑SCLC came from ADRIATIC, a double-blind, phase 3, placebo-controlled trial. ADRIATIC compared durvalumab monotherapy and placebo in people with stage 1 to 3 LS‑SCLC who had had 4 cycles of first-line cCRT without disease progression and had an Eastern Cooperative Oncology Group status of 0 or 1. At the January 2024 data cut-off, median overall survival was 55.9 months for durvalumab and 33.4 months for placebo (stratified hazard ratio [HR] 0.73, 98% confidence interval [CI] 0.54 to 0.98). Median progression-free survival was 16.6 months for durvalumab and 9.2 months for placebo (stratified HR 0.76, 97% CI 0.59 to 0.98). The EAG noted that all people in the trial had tumour assessments at specified intervals and considered that this adequately represented active monitoring. The EAG's clinical experts advised that the improvement in median overall survival seen with durvalumab is clinically meaningful. The committee concluded that the trial data showed that durvalumab improves overall and progression-free survival compared with placebo.

The company's economic model was a partitioned survival model with 3 health states: progression-free, progressed disease and death. Health state occupancy over time was informed directly by ADRIATIC overall survival and progression-free survival extrapolations (see section 3.6). All people entered the model in the progression-free health state, and could transition to progressed disease or death. Patients in the progressed-disease health state were also at risk of transitioning to death. The EAG advised that the company's model structure is appropriate and follows the same structure used in previous NICE technology appraisals for lung cancer treatments. The committee agreed the company's model was acceptable for decision making and the modelling and analyses had been done to a good standard.

The company fit a range of standard parametric and spline models to survival data from ADRIATIC. The company explained that spline models were selected because they accommodated the complex hazard functions seen in the ADRIATIC data. The company selected the 2-knot spline normal model for extrapolation of overall survival, and the 1-knot spline normal model for progression-free survival. The same model was selected for both treatment arms, but models were fit independently for each arm because the proportional hazards assumption was not assumed to hold. The company also applied a cure fraction of 90% to people who were progression-free 5 years after starting treatment in both arms. The EAG noted that alternative models provided a better statistical fit to data from ADRIATIC. The EAG preferred to use the 1-knot spline hazard model for overall survival, and the generalised gamma model for progression-free survival. The EAG preferred not to apply a cure fraction separately, because it noted that its spline model already accommodated the complex hazard function seen in the trial data. It also noted that people whose cancer does not relapse after 5 years may still have long-term toxicity from radiotherapy. The clinical experts explained that after 5 years without progression the chance of the cancer coming back is very low, but it can in rare cases. The committee noted that most people having durvalumab for LS‑SCLC would not have had surgical resection, and durvalumab would be used as a maintenance treatment. So, most people would still be living with cancer after initial treatment. The clinical experts considered the 10‑year survival estimates from the EAG's and company's extrapolations and advised that both are plausible, but uncertain. The committee noted that the choice of progression-free survival curve had minimal impact on the incremental cost-effectiveness ratio (ICER), but choice of overall-survival curve did have an impact. Removing the cure fraction had minimal impact on the ICER. Although both the company's and EAG's extrapolations for overall survival were plausible, the EAG's extrapolation for durvalumab appeared to be a better visual fit to the Kaplan–Meier data. As with the treatment effect waning assumption (see section 3.7), the committee agreed that the EAG's extrapolations were able to capture the hazards at the end of observed follow-up period without the need for any further adjustments. So, the committee preferred the EAG's base case that used the 1-knot spline hazard model for overall survival and the generalised gamma model for progression-free survival, without a cure assumption applied.

According to the marketing authorisation, durvalumab should be stopped on disease progression, unacceptable toxicity or after 24 months (2 years). The company's model did not include treatment effect waning. The company stated there was no evidence to support this. The EAG noted that the median overall survival follow-up period in ADRIATIC for durvalumab (30.75 months) may not have been long enough to rule out treatment effect waning, which may occur far beyond the end of trial follow up. So, the EAG explored several scenarios to explore the impact of including treatment effect waning. It explained that the most plausible of these involved equalising the hazards in both arms. It presented scenarios equalising the hazards between arms immediately at 5 years after starting treatment and gradually between 3 and 5 years after starting treatment. The EAG explained that although there is no evidence for treatment effect waning there is uncertainty about the biological plausibility that after treatment is stopped, it still has an impact on the disease. The clinical experts explained that immunotherapies can have a sustained benefit, which continues for some time after stopping treatment. They explained the difficulties in predicting survival at 10 or 20 years, including how treatment effect waning may influence these estimates. The committee noted that applying treatment effect waning had a large impact on the ICER particularly in the EAG's base case, which applied waning to the full cohort because a cure assumption was not applied. The committee agreed that the decision to apply treatment effect waning should be based on evidence specific to this evaluation, including the follow-up data available and the selected survival extrapolations. The committee decided that, given the fixed treatment duration of 2 years, the overall-survival follow-up duration in ADRIATIC was sufficient despite the EAG's concerns. It noted that the trial data already appeared to capture some increase in hazards after people stopped treatment at 2 years and that spline models were flexible enough to capture the trajectory of the observed data. The committee agreed that some people have durable responses to immunotherapies that are maintained after treatment is stopped. It noted that this can be seen more clearly when immunotherapies are used in early-stage cancers. Rates of relapse may increase after durvalumab is stopped, but the extent to which this occurs should already be captured in the observed data and the survival extrapolations. So, the committee concluded that it preferred to use the available data and not apply an additional adjustment for treatment effect waning.

The company's estimates for resource use were taken from NICE's technology appraisal guidance on durvalumab for maintenance treatment of unresectable non-small-cell lung cancer after platinum-based chemoradiation. This was based on data from the PACIFIC trial. The EAG made the following adjustments to the company's estimates based on clinical expert opinion:

• including outpatient oncologist visits during durvalumab treatment

• including CT scans instead of chest X‑rays

• assuming a higher number of blood tests for people with progressed disease.
  
  The total resource-use cost per year in the EAG's base case was higher than in the company's base case, mainly driven by the frequency of outpatient visits. During the committee meeting, the company explained it had not included costs for outpatient visits during durvalumab treatment because it considered this to be double counting. The company assumed that the cost for these visits would be captured by the administration costs already applied. The clinical experts explained that people having durvalumab would see an oncologist separately to having durvalumab administered, so the cost of oncologist visits should be included in addition to the cost of administration. They also explained that CT scans rather than chest X‑rays would be used, because chest X‑rays are not used for ongoing monitoring. They advised that monitoring after progression was more likely to be in line with the EAG's higher estimates because people may go on to have subsequent treatments. The committee concluded that the EAG's estimates of resource use were more aligned with expected resource use in clinical practice.

The company's model only included the costs for subsequent (second-line) treatments. The company stated this was because the effects were already captured by post-progression survival data from ADRIATIC. The types and proportions of subsequent treatments in the company's base case were based on ADRIATIC data, adjusted based on clinical expert input at an advisory board. The EAG's clinical experts advised that subsequent treatment choice depends on how quickly the cancer relapses. The EAG explored estimates based on clinical expert opinion as a scenario. But in its base case, the EAG preferred to use unadjusted subsequent treatment data from ADRIATIC to align costs and effects. The committee noted the very small impact of this on the cost-effectiveness estimates and agreed that the EAG's preferred approach of aligning costs and effects was appropriate.

NICE's manual on health technology evaluations notes that, above a most plausible ICER of £20,000 per quality-adjusted life year (QALY) gained, judgements about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also take into account other aspects including uncaptured health benefits.

The committee considered that, if approved, durvalumab would be the first treatment available in this indication and address an unmet need. It noted that the consequences of decision error if durvalumab was recommended were low, based on the small number of people expected to be eligible for treatment in clinical practice. It also noted that the length of follow up in the trial reduces the uncertainty about the duration of treatment benefit after stopping durvalumab, but the size and duration of the long-term benefits are still uncertain. So, the committee concluded that an acceptable ICER would be around or slightly above the middle of the range NICE considers a cost-effective use of NHS resources (£20,000 to £30,000 per QALY gained).

The committee did not identify any equality issues that may arise from its recommendation.

The committee considered whether there were any uncaptured benefits of durvalumab. It did not identify additional benefits not captured in the economic modelling. So, the committee concluded that all additional benefits of durvalumab had already been taken into account.